{ "schema_version": 1, "builder_plan": { "scenario": "Long customer conversations exhaust memory or limit concurrent requests.", "change": "Change only the attention cache policy, after checking model and runtime support.", "baseline": "The same model with its ordinary attention cache.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Peak cache memory, long-context answer quality, first-token latency, and requests completed.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2610.31098", "title": "RIFT: KV Cache Compression for Draft Models with Complementary Compressibility Training", "summary": "RIFT replaces static prefill-time KV cache selection in speculative drafters with periodic selection refreshes during decoding, plus a sparse-mask training phase that makes the drafter robust to a compressed cache. The costs are extra decode latency from refresh score calculations (partially mitigated by an adaptive refresh schedule) and the engineering and training burden of fine-tuning the drafter for compressibility. It can fail where acceptance-length gains do not offset refresh latency, where the adaptive schedule misjudges refresh timing on unfamiliar workloads, or where the reader lacks the pipeline to retrain and validate a drafter against their target model.", "rationale": "The method targets a real constraint (drafter KV memory in long-output speculative serving on a single GPU), but its benefit depends on the reader's acceptance-length profile and their ability to retrain a drafter, so it warrants a small-scale pilot rather than direct adoption.", "technique": "Refreshed KV cache selection for speculative drafters", "family": "cache_kv", "recommendation": "testar", "published": "2026-10-22", "as_of": "2026-10-23", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.31098", "independent_implementations": 0, "total_implementations": 1, "stars": 122, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "RIFT replaces static prefill-time KV cache selection in speculative drafters with periodic selection refreshes during decoding, plus a sparse-mask training phase that makes the drafter robust to a compressed cache. The costs are extra decode latency from refresh score calculations (partially mitigated by an adaptive refresh schedule) and the engineering and training burden of fine-tuning the drafter for compressibility. It can fail where acceptance-length gains do not offset refresh latency, where the adaptive schedule misjudges refresh timing on unfamiliar workloads, or where the reader lacks the pipeline to retrain and validate a drafter against their target model.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.31098", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "RIFT with compressibility training retains over 98% of full-cache acceptance length on 32K-token inputs and 4K-token outputs while holding the drafter KV cache to an 8K-token budget (roughly a 4x reduction relative to the full context).", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.31098", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_fetched", "core_kind": "", "equations_fetched_at": "", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }